VideoCC3M

该数据集包含了大约250万个视频-文本对，这些对是通过从CC3M中挖掘与种子图像相似的视频构建而成的。此外，该数据集还用于评估ReSpec过滤方法的有效性。其规模大约为250万个视频-文本对，适用于零样本视频检索任务。

This dataset comprises approximately 2.5 million video-text pairs, which are constructed by mining videos similar to seed images from CC3M. Furthermore, this dataset is employed to evaluate the effectiveness of the ReSpec filtering method. With a scale of around 2.5 million video-text pairs, it is suitable for zero-shot video retrieval tasks.